MuxLayer

The local model control layer for coding agents.
Use Codex, Claude Code, Gemini CLI, and OpenCode with the models you actually want.

Codex · Claude Code · Gemini CLI · OpenCode
↓
MuxLayer
↓
DeepSeek · Kimi · MiMo · OpenAI · Anthropic · OpenRouter · Ollama · 25+ model providers

中文 · Download · Quick Start · Full Reference · 💬 Discussions

MuxLayer is a local gateway for AI coding workflows. Requests enter through one controlled endpoint, then MuxLayer routes, converts, passes through, fails over, and records what happened.

Why MuxLayer

• Keep your clients: Continue using Codex, Claude Code, Gemini CLI, OpenCode, and AtomCode with one-click config restore.
• Choose the model path: Route each request to the provider and model you want, with protocol conversion when needed.
• See every request: Inspect route decisions, payload diffs, upstream errors, tokens, cost, latency, and failover locally.

MuxLayer is not a hosted API reseller or a generic proxy. It is a local control layer between your coding agents and the model APIs they use.

Quick Start

1. Download MuxLayer and open it.
2. Open Quick Setup or Providers, then paste a provider API key.
3. Start the gateway. The default endpoint is http://127.0.0.1:9090.
4. Open Clients and apply the Codex, Claude Code, OpenCode, Gemini CLI, or AtomCode configuration.

Send a test request, then open Logs to see the selected provider, model, route, status, and latency. The original client configuration can be restored at any time.

What it handles

Capability	What it gives you
Protocol conversion	Responses, Chat Completions, and Anthropic Messages can meet the upstream protocol they need.
Routing and failover	Select providers by route profile, conditions, health, cooldown, and fallback order.
Request tracing	Compare raw and converted payloads, route decisions, upstream status, latency, tokens, and cost.
Session affinity	Keep multi-turn conversations on the provider that started them, unless an override is required.
Local models and budgets	Discover Ollama / LM Studio endpoints and optionally gate daily spend.

Screenshots

Request trace	Overview

Supported clients and providers

Clients: Codex · Claude Code · Gemini CLI · OpenCode · AtomCode · Cursor / Continue / Cline

Providers: OpenAI · Anthropic · DeepSeek · Kimi · MiMo · Gemini · OpenRouter · Groq · Mistral · Ollama · LM Studio · and 25+ model providers.

See the full provider compatibility matrix and the usage guides.

All supported providers

Provider	Type	Native Protocols	Provider-Specific Handling
Xiaomi MiMo	mimo	Chat + Anthropic	Multi-turn reasoning_content round-trip, region-aware tp-* host auto-routing, temperature strip in thinking mode, tool_choice non-auto strip, omni web_search strip, web_search builtin gated by matrix, Web Search Plugin auto-degrade / retry
DeepSeek	deepseek	Chat + Anthropic	Vision model preserves image inputs; text-only models strip images with an explicit notice; DeepSeek V4 thinking history reasoning backfill, schema cleaning, message reordering
Anthropic (Claude)	anthropic	Anthropic	tool_use/tool_result, input_schema, thinking budget, native cache_control
GitHub Copilot	copilot	Chat + Anthropic	GitHub token → Copilot bearer exchange, x-initiator billing classification, Claude model dash→dot normalization
OpenAI	openai	Chat + Responses	None (Responses passthrough or Chat conversion)
Google Gemini	google_gemini	Chat	None
Kimi / Moonshot	kimi	Chat + Anthropic	web_search → builtin_function/$web_search; K3 uses reasoning_effort:max (no K2 thinking param); coding models keep thinking on/off control
MiniMax	minimax	Chat	Strip reasoning_effort / response_format, <think> extraction
GLM (Zhipu)	glm	Chat	Generic
DashScope (Qwen)	dashscope	Chat	Generic
SiliconFlow	siliconflow	Chat	Generic
Volcengine (Doubao)	volcengine	Chat	Generic
Baichuan	baichuan	Chat	Generic
StepFun	stepfun	Chat	Generic
SenseNova	sensenova	Chat	Drops null strict / response_format / non-function tools, merges system messages
Yi (01.AI)	yi	Chat	Generic
ModelScope	modelscope	Chat	Generic
xAI (Grok)	xai	Chat	Generic
Mistral	mistral	Chat	Generic
Groq	groq	Chat	Generic
Together	together	Chat	Generic
Fireworks	fireworks	Chat	Generic
Cerebras	cerebras	Chat	Generic
Perplexity	perplexity	Chat	Generic
Cohere	cohere	Chat	Generic
OpenRouter	openrouter	Chat	None
Custom	custom_openai_compatible	Chat	None (set Base URL yourself)

Install

Platform	Package
macOS Apple Silicon	MuxLayer 2.0.1
macOS Intel	MuxLayer 2.0.1
Windows 10 / 11	MuxLayer 2.0.1
Debian / Ubuntu	MuxLayer 2.0.1
Other Linux distros	MuxLayer 2.0.1

On macOS:

brew install --cask dengmengmian/tap/muxlayer

Windows: Builds are currently unsigned, so SmartScreen may warn on first install. See the installation and build notes.

Existing installations using the legacy agentgate cask remain supported and continue receiving the MuxLayer app. New installations should use muxlayer.

Documentation

• Full reference
• Codex Desktop with third-party APIs and plugins
• Codex + DeepSeek
• Codex + Xiaomi MiMo
• Claude Code + DeepSeek
• Claude Code + GitHub Copilot
• Gemini CLI
• OpenCode
• Local models

Open-source toolkit

MuxLayer is part of a small AI developer toolkit:

• CodeLeveler — inspect, edit, run, and verify code in the terminal.
• ReviewGate — review code changes and surface high-confidence issues.

Development

pnpm install
pnpm tauri dev

pnpm brand:check
pnpm test:run
pnpm build

Community

• Questions and setup help: Discussions
• Bugs and provider requests: Issues
• Contributing: CONTRIBUTING.md

Compatibility

MuxLayer was formerly AgentGate. Existing desktop update identity, data directories, and legacy headless names remain compatible; see the brand migration plan.

License

MIT